Learn

AI hardware guide

What the different classes of AI hardware are, and which problems each one exists to solve.

Accelerators

The processor doing the AI maths. Sold as desktop cards, workstation cards or server modules that cannot go in a desktop.

Workstations

A complete deskside system, built and tested around one or two accelerators.

GPU servers

Rack or tower systems holding several accelerators, shared across a team or an application.

Integrated systems

Compute, networking, storage and software delivered and supported as one platform.

Clusters

Multiple servers connected by a dedicated fabric for distributed work.

Rented compute

Any of the above, accessed without owning the hardware.